Platform · Hub

The box that runs the boat.

The hub is the physical box on the boat. A Mac mini sitting in the nav station, connected to the boat's instrument network through a standard NMEA-2000 gateway. It runs the 32°N operating system, hosts the apps, and handles the data layer. It keeps working when the internet is gone.

I picked the Mac mini because it is the right piece of hardware for this job: silent, low-power, thermally stable, and powerful enough to run local AI inference alongside everything else. Off-the-shelf. No custom PCBs, no marinised housings, no proprietary brackets.

What it does Back to the platform →

═══════════════════════════════════════════════════════════════ -->

Scene 02 · the hub · slide 1 · pair your phone

02 · The hub · slide 1

Hold up your phone. The hub does the rest.

Plug the hub in, open the 32°N app, and bring them close. Bluetooth pairs the hub to your account in seconds — no codes, no QR scans, no manual setup. The same handshake feeling as a new iPhone.

SETUP · STEP 1 OF 6 Bluetooth pairing · proximity discovery

Scene 02 · the hub · slide 2 · setting up excess 14

02 · The hub · slide 2

The hub knows your boat.

Once paired, the hub belongs to one boat — Excess 14, Unsinkable II. Your 32°N account, the hub, and the boat are linked from this moment on. Every device you bring aboard will join automatically.

SETUP · STEP 2 OF 6 account & boat linked · Excess 14 · Unsinkable II

Scene 02 · the hub · slide 3 · discover instruments

02 · The hub · slide 3

It finds every instrument on the boat.

The hub speaks NMEA-2000. Plug it into the backbone and it sweeps the bus for chartplotters, AIS receivers, autopilots, wind, depth, GPS — everything that's already on Excess 14. You don't name a single one.

SETUP · STEP 3 OF 6 NMEA-2000 sweep · finds every instrument on the bus

Scene 02 · the hub · slide 4 · network audit

02 · The hub · slide 4

It audits the boat's network.

Every device the hub finds, it tests — default passwords, weak ciphers, exposed services, firmware age. The first marine security tool that already knows your boat. Where there's a fix path, it's one tap away.

SETUP · STEP 4 OF 6 network audit · shield + 1 amber row · marine security baked in

Scene 02 · the hub · slide 5 · device discovery

02 · The hub · slide 5

Every device on board.

Five transports, eleven devices, one screen. The hub speaks NMEA-2000 to your instruments, Bluetooth to your phone, WiFi to your displays, USB to serial gear, and Matter to cabin lights and climate. Everything live, one place — and one row flagged for attention.

SETUP · STEP 5 OF 6 device discovery · 5 transports · 11 devices · 1 attention

What the hub is

Hardware plus adapters.

The hub is two things physically: the Mac mini, and the adapter dongles plugged into it. The Mac mini is the compute. The adapters are the translators — they sit between the Mac mini and the boat's existing NMEA-2000 instrument network.

Boat-class compute

Mac mini M-series. Silent, low-power — around 6 to 10 watts at idle. Thermally stable in a nav station. Native USB-C and Thunderbolt for the full range of NMEA-2000 gateways. Powerful enough to run local AI inference alongside everything else.

Standard NMEA-2000 gateway

One USB gateway between the Mac mini and the boat's instrument backbone. Actisense NGT-1, Digital Yacht iKonvert, Yacht Devices YDNU-02 — all work. 32°N reads the raw PGN frames directly into its own native bus. No third-party server sitting in between.

Hardware, not software

The hub is hardware. The OS runs on the hub. Apps run inside the OS. Get the hardware right and the rest of the platform installs cleanly on top. I do not have a proprietary box — the Mac mini is something you can buy at any Apple Store.

Why a Mac mini

The right tool for this job.

I looked at the alternatives. The Raspberry Pi 5 is capable and cheap. An N100 mini-PC is a reasonable middle ground. The Mac mini costs more. Here is why it is still the right answer for the first version of this.

Power and thermals

The M-series Mac mini idles around 6 watts and peaks around 18 watts under a full AI inference load. A Pi 5 peaks at 12 watts but thermal throttles in a warm nav station without active cooling. The Mac mini does not throttle under sustained load. On a boat with a 12V house bank, keeping a 6-watt device running continuously costs roughly 12 amp-hours a day — easy to budget.

The chassis is completely fanless in the idle-to-light-load range, which is most of what it does at anchor. Under AI inference the fan spins up — still quiet by any reasonable standard, inaudible in a closed nav station.

Software ecosystem

The 32°N platform services run as a Docker Compose stack. Docker Desktop on Apple Silicon is stable, well-tested, and straightforward to update. The container images are the same ones that run in the cloud — there is no special "hub build" that diverges from the rest of the codebase.

Local AI inference runs via Ollama on Apple Silicon. The Neural Engine on M-series chips delivers real-time inference speeds for 7B and 13B parameter models without a GPU. That is what makes on-board, offline AI practical.

Connectivity

Two USB-C / Thunderbolt 4 ports plus two USB-A ports cover the full range of NMEA-2000 USB gateways, SDR receivers, and BLE dongles. The built-in Gigabit Ethernet goes to the boat's network switch. WiFi 6 covers tablet and phone connectivity when you do not want cables. HDMI out to the chartplotter display or a second monitor.

Supply and longevity

Mac minis are sold in every Apple Store, serviced globally, and Apple supports them for a long time — M1 units from 2020 still receive full OS updates. When you are three years into an offshore passage on a budget, that matters more than you think. Spare units are easy to source, and the software state is fully reproducible from the repo.

Adapter catalogue

The gateways I have tested.

A gateway sits between the Mac mini and your boat's NMEA-2000 backbone. All of the ones below surface as a serial device that the 32°N bus service reads directly. The USB options are the most reliable; the WiFi options give you the flexibility of wireless installation.

Actisense NGT-1

The reference USB gateway. Reliable, widely supported, reads the full range of NMEA-2000 PGN frames. Plugs into the NMEA-2000 backbone with a standard Male Type A connector. Powered from the bus. Has been on the market for years and will keep working.

USB · Primary recommendation

Digital Yacht iKonvert

Another solid USB gateway. A slightly lower profile than the NGT-1. Uses the same underlying PGN decoder in 32°N. If you already have one on board, it works without changes.

USB · Tested

Yacht Devices YDNU-02

USB NMEA-2000 gateway from Yacht Devices. Compact, well-made. Compatible with the 32°N bus service and produces clean PGN output across the full range of instrument types.

USB · Tested

Yacht Devices YDWG-02

WiFi gateway — useful when a USB cable run to the nav station is not practical. Broadcasts NMEA-2000 frames over UDP on the boat's WiFi network. 32°N reads from the same PGN decoder regardless of whether the frames arrive over USB or UDP.

WiFi · Tested

Actisense W2K-1

WiFi NMEA-2000 gateway from Actisense. Same wireless installation story as the YDWG-02. If you have an existing Actisense setup and want to go wireless, this is the straightforward choice.

WiFi · Compatible

Other sources — BLE, SDR, MATTER

Victron battery monitors speak Bluetooth Low Energy and feed through 32°N's BLE ingestion service. RTL-SDR receivers add AIS and weather satellite decoding. MATTER devices slot in via the hub's Matter controller. All of them emit the same typed bus events as the NMEA-2000 adapters.

BLE · SDR · MATTER

Where the hub fits

The foundation of the on-boat stack.

The hub is the bottom of the on-boat stack. The bus runs on it. The OS runs on it. The apps run on it. The cloud layer connects out from it. Remove the hub and the rest of the on-boat platform does not have anywhere to run.

What runs on the hub

The OS and the bus.

The hub runs two things: the 32°N OS and the native bus. They are separate concerns. The bus handles the data layer. The OS handles everything else: identity, app lifecycle, the AI gateway, the audit log, and sync.

The 32°N OS

The OS runs as a Docker Compose stack. It starts at boot and manages the platform services: the identity service, the local Postgres database, the event bus, the AI gateway, the audit log, and the sync engine. Apps get all of that by targeting the platform contracts — they do not wire it up themselves.

Security is baked in. Every device that connects to the hub goes through a registration step and gets scoped to specific permissions. The audit log is structured and queryable. Nothing connects that has not been explicitly admitted. The OS itself is signed and updates automatically from the public registry — same as any other 32°N service.

The OS does not depend on Apple Silicon architecturally. It runs on the Mac mini today because that is the right hardware for v1. Other platforms come later.

The native bus

Every sensor reading, state change, and event goes through the 32°N native bus as a typed, timestamped, source-tagged message. Apps subscribe to the bus. They do not poll hardware directly. They do not need to know what kind of GPS is on board.

This is not SignalK. It is not a third-party message broker bolted on later. It is a schema designed from the start to be the thing all the other layers depend on. The schema is open and versioned in the public repo.

Communications between the hub and connected devices — phones, tablets, remote clients — also go through the bus. One bus, one data model, many consumers.

When the internet is gone

The hub keeps working.

The cloud layer is where capability expands, not where the product breaks down. Lose the internet and the hub still does everything that matters on passage.

Works without the internet

Chartplotter, AIS, depth, wind, GPS — all live on the bus and available to every app regardless of connectivity. Weather routing uses the last downloaded model. Anchor watch runs locally. The logbook writes to the local database. AI inference runs on-device via Ollama — no cloud round-trip.

Picks up when you reconnect

The sync engine runs a queue. When the hub reconnects — marina WiFi, satellite link, phone hotspot — it catches up with the cloud: new weather models downloaded, logbook synced, any pending AI tasks processed. No manual intervention needed.

Safety surfaces always work

Anchor watch, collision alarms, depth alarms, engine temperature warnings — these are all local. They run on the hub bus, trigger alerts through the hub, and do not need a cloud round-trip. Safety is not a cloud feature.

No degraded-mode UX

The apps do not show "offline mode" banners or disable features when the internet is gone. They show what they have. If the latest weather model is two days old, it says so. The rest keeps working.

Security model

Explicit admission, not implicit trust.

Every device that connects to the hub is registered, scoped, and logged. There is no "anything on the boat WiFi can talk to the hub" shortcut. The audit trail is structured and queryable.

Surface	How it is secured
Device registration	Every client device — phone, tablet, laptop — registers with the hub and receives a scoped token. Tokens are revocable. The hub maintains a device registry in the audit log.
App sandbox	Each app runs in its own container with access only to the bus topics it has declared. Apps cannot reach each other's data stores directly.
AI gateway	AI agents access boat data through typed MCP tools, not raw database access. Each tool call is capability-checked and logged. Agents cannot escalate to arbitrary data access.
Audit log	Structured Postgres log, queryable, tamper-evident. Every state-changing operation — waypoint set, alarm acknowledged, device admitted — writes a record. Used by the security scanner.
Updates	Platform services and apps update from signed container images. The update channel is the same public registry as the open-source repo. No silent updates; the hub logs every change.

Getting started

From hardware to running in an afternoon.

The installation is designed to be done by someone comfortable setting up a NAS or a home server. You do not need to be a programmer. You do need to be comfortable at a terminal.

Get the hardware

Mac mini M2 or M4. 8 GB RAM is the minimum; 16 GB is better if you are running local AI inference. Plus a USB NMEA-2000 gateway — Actisense NGT-1 is the simplest choice if you are starting from scratch.

Install Docker Desktop

Docker Desktop for Apple Silicon. Free for personal use. This is the runtime that the 32°N platform services run on.

Clone the repo and run the compose stack

One git clone and one docker compose up. The compose file pulls the signed service images. On first run it initialises the database, generates the hub identity keypair, and starts the bus service.

Connect the NMEA-2000 gateway

Plug the USB gateway into the Mac mini. Plug the gateway's NMEA-2000 Male Type A connector into your boat's backbone. The 32°N bus service detects the serial device and starts reading PGN frames within a few seconds.

Open the dashboard on any browser on the boat network

The hub exposes a local web UI on port 3000. Open it on your phone, tablet, or any laptop on the boat WiFi. Register the device, and you have access to the full platform — instruments, apps, AI assistant, everything.

Scene 02 · the hub · slide 6 · plan your systems

02 · The hub · slide 6

Plan your systems.

Once the hub is on board, plan your boat with it. Size the battery bank. Spec the solar. Map the bus. Model what you need before you buy a single connector — the planning lives inside the hub, not in a spreadsheet on shore.

See it in the demo, or read about the rest of the platform.

The demo lets you explore the chartplotter and cockpit instruments without any hardware. The platform page has the full five-layer architecture and links to every sub-page.

Back to the platform →